There is no valid evidence that genes play a role in causing ‘schizophrenia’

In a 2003 edition of Science, the editors declared that the identification of genes for “mental illness” was the second most important “scientific breakthrough” of the year.

“Schizophrenia, depression, and bipolar disorder often run in families,” the editors wrote, “but only recently have researchers identified particular genes that reliably increase one’s risk of disease.” The prefrontal cortex, they wrote, “is regulated in part by a gene called COMT, one of the handful associated with risk of schizophrenia”.

As it turned out, claims based on these “particular genes” didn’t hold up, including the COMT gene. When a journal as prestigious as Science says that genes for psychiatric conditions have been discovered, most people take this as fact — and remember it as fact. But in this case, it turned out not to be a fact. Despite the development of new techniques leading to current claims of gene associations based on genome-wide association (GWAS, based on common single nucleotide polymorphisms, or SNPs), polygenic risk score (PRS), and copy number variant (CNV) studies, genes shown to cause “schizophrenia” — a questionable concept in and of itself — remain undiscovered.

The criteria for diagnosing schizophrenia remain vague and subjective. As psychologist John Read has shown, there are 15 ways that two people can meet the DSM criteria for schizophrenia without sharing any symptoms in common. Reliability remains low, and is even decreasing. I agree with writers who reject the “schizophrenia” concept and prefer the term “psychosis.”  However, as critical psychiatrist R. D. Laing once remarked at an academic conference, “I am unhappy about using the term schizophrenia at all. But it would be somewhat whimsical to eliminate it from my vocabulary, since it is on the lips of so many”.

I became interested in the “genetics of schizophrenia” topic as a clinical psychology graduate student in the mid-1990s. The arguments fascinated me, and because I saw the genetic argument as weak, it was astonishing to hear that the debate had been largely closed in favour of genetics two decades earlier. I decided to focus my 1998 doctoral dissertation on a critical analysis of schizophrenia genetic research. I have been writing about genetic research in the social and behavioural sciences ever since, including my 2015 book The Trouble with Twin Studies.

In this article, I make the case that, contrary to most of what has been written on the topic, there exists no valid evidence that genes play a role in causing schizophrenia. I make the complete case in the 2023 Routledge English-language edition of my book Schizophrenia and Genetics: The End of An Illusion (a greatly revised version of a previously self-published e-book). Here I summarize the book’s main arguments in favour of reopening the genetics of schizophrenia debate.

Family, twin and adoption studies

Science is about data, but more importantly about how scientists and others interpret data. The common claim that schizophrenia is “80% heritable,” which both justifies and guides the search for genes at the molecular genetic level, is based on a century of family, twin, and adoption research. Because similar methods are used in other areas of psychiatry and in behavioural research in general, the critique of these methods has major implications for most areas of human behaviour. The production/usefulness of heritability estimates (which range from 0% to 100%) is controversial in its own right, which calls into question statements of “percentages of variance in disease liability” often used to support causal models of psychiatric conditions.

It is widely agreed upon that family studies are unable to disentangle the potential influences of genes and environment on human behavioural differences. Twin studies assume that reared-together MZ (identical) and DZ (fraternal) twin pairs grow up experiencing “equal environments,” and that the only behaviourally relevant factor distinguishing these pairs is their differing degree of genetic relationship to each other (100% versus an average 50%). Critics and even most genetic researchers however, recognize that MZ environments are more similar than DZ environments. Although twin researchers and their supporters have developed various arguments and tests in attempts to sidestep twin studies’ glaring unequal environments problem, I show in Schizophrenia and Genetics that these arguments and tests do not hold up. Twin studies’ crucial MZ-DZ “equal environment assumption” (EEA), therefore, is false. Publications based on this obviously false assumption have produced some pretty strange “findings,” including that dog ownership, vegetarianism, and concern for nature are only somewhat less heritable than schizophrenia.

Because family studies and twin studies are unable to disentangle the potential influences of genes and environments on human behavioural differences, they supply no evidence that genes play a role in causing schizophrenia. Seventeen schizophrenia twin studies have been published since 1928. Since that time, leading authorities in twin research, psychiatry, and other behavioural fields have mistakenly attributed to genetic factors the usual finding of greater MZ versus DZ behavioural resemblance (concordance).

The famous Danish-American schizophrenia adoption studies were flawed and biased to the point of constituting a not-yet-recognized scientific scandal (more below), and genetic interpretations of the few remaining adoption studies are invalid due to bias, methodological issues, and environmental confounds.

To summarize, family, twin, and adoption studies have failed to produce evidence that schizophrenia is “heritable”.

Environmental factors

Meanwhile, studies have linked schizophrenia and psychotic conditions to childhood adversities such as having experienced bullying, emotional abuse, incest, neglect, parental loss, physical abuse, and sexual abuse, as well as the stresses of immigration, urbanicity, poverty, relative poverty, and social marginalization. Brain disease theories survive through the use of smoke and mirrors and the confusion of cause and effect. From the perspective of those who see political, economic, social, and oppressive aspects of society as causing widespread psychological harm, what psychiatry calls the “societal burden of mental disorders” is really the mental burden of societal disorders.

Molecular genetic research

The mistaken belief that family, twin, and adoption studies long ago established schizophrenia as a genetically based disease constitutes the fundamental error of schizophrenia gene-finding strategies. Neuroscientist/geneticist Kevin Mitchell wrote in 2009, “Familiality and twin concordance data are the bedrock on which all psychiatric genetics, including GWAS, is based and justified”. What Mitchell called bedrock is more like quicksand.

Corporate media reporting of false-alarm “schizophrenia gene discoveries” has a long history, and continues to the present day. A widely reported yet non-replicated schizophrenia gene association was published in 1988 by the Sherrington group, who believed they had found “the first strong evidence for the involvement of a single gene in the causation of schizophrenia.” A November 10, 1988, front-page New York Times article about this study proclaimed, “Schizophrenia Study Finds Strong Signs of Hereditary Cause”.. The Times also reported subsequently non-replicated schizophrenia gene discoveries in 1995, 1997, 2002, 2006, 2008, and 2013, with headlines such as “Brain-Tied Gene Defect May Explain Why Schizophrenics Hear Voices,” “Schizophrenia May Be Tied To 2 Genes, Research Finds,” “Schizophrenia as Misstep by Giant Gene,” and “Study Ties Genetic Variations to Schizophrenia.” Retractions are few and far between.

Linkage and candidate gene studies

Despite much fanfare and anticipated gene-discovery “euphoria” dating back to the 1970s, the schizophrenia linkage and candidate gene eras eventually crashed and burned. In a 2002 analysis, leading schizophrenia molecular genetic researcher Lynn DeLisi and colleagues concluded, “No linkage appears to be consistently replicable across large studies”. In his 2018 book Blueprint: How DNA Makes Us Who We Are, top behavioural geneticist Robert Plomin described over two decades of behavioural candidate gene research as an “approach [that] failed everywhere,” a “fiasco,” and a “flop”. By Plomin’s 2018 tally, for schizophrenia alone “over 1,000 papers reported candidate gene results for more than 700 genes”. Plomin then asked, “how can so many published papers have got it so wrong?” Factors such as a reliance on twin studies and heritability estimates, systematic bias, and the use of the questionable research practices I will soon discuss provide a partial explanation. Yet leading researchers (including Plomin), review writers, and the authors of influential books told the public in the 1990s and 2000s that studies had identified genes associated with schizophrenia. Similar stories of claims and hype followed by a spectacular failure also played out in the areas of major depression and intelligence (IQ).

Commenting in 2008 at the height of the failed candidate-gene era, psychiatric researcher Timothy Crow wrote that although the “schizophrenia gene pond” was in fact “empty”, a “chorus of reviews…pervades the literature and will convince all but the most inquisitive that a solid foundation of evidence supports the pathophysiological relevance of these candidate genes”. When evaluating current claims of GWAS gene associations or discoveries by the world’s top researchers, we must remember that many of them were claiming gene discovery in an era they now recognize to have been a “flop”. We should keep Crow’s 2008 “chorus of reviews” comment in mind when assessing recent claims based on newer methods — same chorus, slightly modified lyrics.

A 2016 New York Times article with the title “Scientists Move Closer to Understanding Schizophrenia’s Cause” reported on a supposed discovery that variants of the C4 gene, which codes for a protein in a cascade of factors in the innate immune system and synaptic sculpting, play a role in causing schizophrenia. In 2022 this widely publicized C4-schizophrenia link was shown to be false, as another psychiatric genetic “landmark study” failed to hold up as the schizophrenia genetics story continues to unravel. (The 2022 Borbye-Lorenzen et al. study failing to find an association between C4 and schizophrenia was published as I was writing this article, too late to be included in the book.)

Enter GWAS and PRS studies

Many current gene association claims are based on the “hypothesis-free” genome-wide association (GWAS) method, which dates back to the period 2005-2007. By definition, a scientific fishing expedition is a hypothesis-free method, where researchers base their conclusions on significant yet chance associations that in the GWAS context can pop up on a Manhattan plot. Schizophrenia GWAS publications, which can be seen as describing fishing expeditions of this type, have produced only spurious or non-causative “gene associations” (correlations). Correlation, as is well known, does not imply cause. In 2022, Thomas Insel, the biologically oriented former director of the U.S. National Institute of Mental Health, recognized that “in contrast to the mutations discovered for cancer or rare diseases, none of the genetic variants associated with mental illness can be considered causal”.

In Blueprint, Plomin described the polygenic risk score (PRS) method as a molecular genetic technique that combines statistically significant and nonsignificant individual SNP associations identified in a GWAS to produce a polygenic (composite) risk score. A 2022 study by the Schizophrenia Working Group of the Psychiatric Genomics Consortium calculated a 7% schizophrenia polygenic risk score. The study’s lengthy “Competing Interests” section showed that many of the researchers taking part in this project had financial ties to the drug companies.

In addition to the above-mentioned problems, the GWAS and PRS methods are subject to confirmation biases, environmental confounds such as population stratification (“popstrat”), financial conflicts of interest, and a reliance on questionable or false assumptions.

CNV research

Another area focuses on potential rare risk variants, such as rare copy number variants, or “CNVs.” Many studies claiming CNV-schizophrenia associations have appeared in recent years, but they are subject to many of the problems found in the GWAS/PRS area. In 2020 Jonathan Flint and Kenneth Kendler, two of the world’s leading psychiatric genetic researchers, wrote that the “early hope that CNVs would reflect the ‘royal road’ to understanding molecular genetic effects on schizophrenia has been disappointing”.

Most likely, Robert Plomin’s final evaluation of the behavioural candidate gene era will be the future consensus evaluation of the psychiatric GWAS/PRS/CNV era as well — approaches that “failed everywhere,” and “fiascos” and “flops” that make us wonder why “so many published papers…got it so wrong”

Misleading information: the National Health Service website

One of countless examples of misleading information about schizophrenia and genetics from a source most people trust is found at the U.K. National Health Service (NHS) website. I will quote from the NHS “Causes — schizophrenia” page as it appeared on 18 November 2022, accompanied by my responses.

• “Schizophrenia tends to run in families, but no single gene is thought to be responsible.”

Response: Schizophrenia running in families can be caused by nature or nurture, as most psychiatric genetic researchers readily acknowledge. And yet, as stated in the Fifth Edition of the DSM (DSM-5), “most individuals who have been diagnosed with [schizophrenia] have no family history of psychosis”. No genes shown to cause schizophrenia have been discovered.

 

• “It’s more likely that different combinations of genes make people more vulnerable to the condition. However, having these genes does not necessarily mean you’ll develop schizophrenia.”

Response: This is speculation, not scientific fact.

 

• “Evidence that the disorder is partly inherited comes from studies of twins. Identical twins share the same genes.”

Response: Twin study MZ-DZ comparisons are confounded by environmental factors, and like family studies the results cannot be interpreted genetically.

 

• “In identical twins, if a twin develops schizophrenia, the other twin has a 1 in 2 chance of developing it, too. This is true even if they’re raised separately.”

Response: Pooled MZ twin concordance in the better-performed studies published since 1963 is less than 25%. But even if it were 50% (“1 in 2 chance”), MZ concordance alone does not establish “heritability.” MZ concordance must be compared with DZ concordance, based on the assumption that MZ and DZ environments do not differ. There are no schizophrenia studies of twins raised separately. Rather, there exist only a handful of similarity-biased anecdotal reports on individual MZ twin pairs who in most cases were only partially reared-apart.

 

• “In non-identical [DZ] twins, who have different genetic make-ups, when a twin develops schizophrenia, the other only has a 1 in 8 chance of developing the condition.”

Response: This result can be completely accounted for by the much less similar environments experienced by non-identical (DZ) versus identical (MZ) twin pairs.

 

• “While this [DZ twin rate] is higher than in the general population, where the chance is about 1 in 100, it suggests genes are not the only factor influencing the development of schizophrenia.”

Response: This result suggests nothing at all about genetic influences, since DZ twins grow up together in the same family and social environment at the same time (birth cohort), and share a twin relationship. For these non-genetic reasons, we would expect the pooled DZ concordance rate to be greater than the 1% general population rate (“about 1 in 100”). The actual pooled DZ rate in the better-performed twin studies is not 12%, as the NHS says (“1 in 8 chance”), but only around 4% (a recent 12/367 3% DZ concordance rate can be found here).

How myths are created and sustained

If the mainstream genetics of schizophrenia story is wrong, the question then becomes why so many people still believe it. This story — as well as the mainstream story of the genetics of most other areas of human behaviour — follows a seven-step process. Twenty-five years of research analysis has led me to conclude that this is a corrupted process, and it goes something like this: academic researchers in the fields of behavioural genetics and psychiatric genetics produce unsound research based on false assumptions and/or manipulated or genetically misinterpreted data, which is then accepted for publication in peer-reviewed academic journals → researchers producing this unsound research are often rewarded, funded, tenured, and even honoured, which motivates them to produce even more unsound research, sometimes over an entire career → respected academic fields and government agencies (including health-related agencies) endorse and promote this unsound research in textbooks, websites, and other publications → the mainstream (corporate) media reports on and promotes this unsound research, often in the form of articles and news reports of new discoveries based mainly on twin research → the mainstream media regularly reports on supposedly exciting new molecular genetic behavioural gene discoveries as if decades of non-replicated false-alarm claims had never happened → books, videos, online articles, and social media posts by journalists and some highly respected researchers and authors promote and celebrate unsound research, while the works of critics are usually ignored, distorted, or dismissed → students and teachers in the academic world, political policy-makers, and the general public are convinced by the above process that what are in fact unsound studies and false-alarm or non-causative behavioural gene association claims are actually sound studies and true causative gene discoveries. This leads them to accept and possibly promote various related political, social policy, scientific, and social-relations viewpoints.

Science’s “replication crisis”

The scientific research/publication process is in a state of crisis, which has much relevance to schizophrenia and other areas of genetic research. It is known as the replication crisis (also known as the “replicability crisis” or the “reproducibility crisis”), meaning a crisis brought about by the discovery that some key findings across various scientific fields were probably non-findings resulting from research that was poorly performed, manipulated to match researcher or funding source expectations, or even fraudulent.

Questionable research practices (QRPs)

Behavioural scientist Leslie John and colleagues introduced the questionable research practices (QRP) concept in 2012. “Although cases of overt scientific misconduct have received significant media attention recently,” they wrote, “questionable research practices (QRPs)…increase the likelihood of finding support for a false hypothesis”. QRPs “are often questionable as opposed to blatantly improper,” and “offer considerable latitude for rationalization and self-deception.” Not just self-deception, I would add, but deception of their colleagues and of society as a whole.

John and colleagues listed ten QRPs: (1) “Failing to report all of a study’s dependent measures,” (2) “Deciding whether to collect more data after looking to see whether the results were significant,” (3) “Failing to report all of a study’s conditions,” (4) “Stopping collecting data earlier than planned because one found the result that one had been looking for,” (5) “‘Rounding off’ a p value,” (6) “Selectively reporting studies that ‘worked,’” (7) “Deciding whether to exclude data after looking at the impact of doing so on the results,” (8) “Reporting an unexpected finding as having been predicted from the start,” (9) “Claiming that results are unaffected by demographic variables…when one is actually unsure (or knows that they [are]),” and (10) “Falsifying data”.

Where we decide to draw the line between QRPs and outright fraud is a matter of opinion. The QRP framework provides an increasingly accepted language describing how and where research goes wrong, even to the point of retraction.

The behavioural science research/publication system remains broken 

A long-running research/publication system that allows people — behind the scenes and prior to submitting manuscripts to journals that usually publish only statistically significant findings — to collect data, analyze results, and reach conclusions influenced by confirmation biases and possible financial conflicts of interest, is an open invitation to engage in QRPs. Plomin recognized in 2022 that the authors of “thousands of” previous behavioural candidate gene reports “commit[ted] every sin in the catalogue of questionable research practices,” an unintended yet blistering indictment of the behavioural science peer-review system that published these reports for two decades. Candidate-gene-era friendly peer reviewers, editors, and journals greenlighted these QRP-riddled studies by the thousands only a few years ago, and there is good reason to believe they continue to do so with potentially QRP-riddled GWAS, PRS, and CNV studies as well.

Among QRPs we find p-hacking, which “occurs when researchers collect or select data or statistical analyses until nonsignificant results become significant” (the conventional level of statistical significance is .05, or the GWAS level of 5 × 10⁻⁸). For example, I have shown that genetic conclusions in the famous Minnesota Study of Twins Reared Apart 1990 Science IQ study were the result of p-hacked research. In molecular genetic research, behavioural geneticist Eric Turkheimer wrote in 2016 that “genome-wide association is unapologetic, high-tech p-hacking.”

Research preregistration

The use of QRPs and the prevalence of p-hacking in psychiatric and behavioural research (including environmental research) would be reduced by research “preregistration,” where investigators would have the option or be required to submit their research rationale, hypotheses, design and analytic strategy, and planned data collection stop-point to a journal for peer review before collecting and analyzing data.

The pre-registration of behavioural research would greatly reduce p-hacking and the use of other QRPs brought to light in the replication crisis. There is growing support for and use of Registered Reports, which as described at the Center for Open Science website,

“is a publishing format that emphasizes the importance of the research question and the quality of methodology by conducting peer review prior to data collection. High quality protocols are then provisionally accepted for publication if the authors follow through with the registered methodology… . It eliminates a variety of questionable research practices, including low statistical power, selective reporting of results, and publication bias, while allowing complete flexibility to report serendipitous findings.”

Behavioural research would benefit greatly by adopting Registered Reports or similar procedures, and the scientific community and the public would have much greater confidence in the results.

Schizophrenia genetic research and the replication crisis

A wise person once said, “Data don’t tell stories, scientists tell stories.” In the current replication crisis era, previously accepted stories researchers told about their data are receiving increasing attention and scrutiny. In the past, critics of psychiatric research didn’t have a widely agreed-upon framework and language to explain how unsound research in this field is performed, published, and validated. The QRP and p-hacking concepts now provide such a framework, enabling those of us with perspectives that differ from mainstream behavioural science positions to tell our stories in new and better ways. In Schizophrenia and Genetics, I show how schizophrenia genetic research has been characterized by the use of QRPs, by a reliance on false or highly questionable assumptions in twin research, adoption research, and other areas, and by spurious or non-causative media-hyped gene-association claims. The paradigm is poised for collapse.

The Danish-American adoption studies

Although still routinely cited in psychiatry textbooks and other works in support of genetics, the famous Danish-American schizophrenia adoption studies (published between 1968 and 1994) provide striking examples of massively flawed research based on several QRPs. Among many other problem areas, to arrive at desired genetic conclusions, investigators Seymour Kety, David Rosenthal, Paul Wender and colleagues (1) dismissed or minimized the impact of environmental confounds such as the selective placement of children by adoption agencies, late separation, late placement, and the non-representativeness and restricted range of adoptive family environments (John and colleagues’ QRP #9); (2) appear to have changed their planned 1968 key group comparison at the last minute because this comparison did not produce statistically significant results in the genetic direction (QRPs #6, #7, and #8); (3) for the same reason, performed a last-minute reduction of their 1994 “Provincial” study index and control adoptee (proband) groups respectively from 42 to 33, and from 42 to 24 (QRP #7); (4) broadened the definition of schizophrenia “as widely as it may have ever been reasonably conceived before” (Rosenthal), and then re-narrowed the definition when necessary, over a 26-year period (QRPs #1, #6, and #7); and (5) temporarily removed a study-defined “schizophrenia spectrum diagnosis” to achieve statistically significant results in one of their key group comparisons (QRPs #1, #6, and #7).

In the Danish-American Rosenthal-led adoption study using diagnosed parents as the first-identified relatives (probands), the researchers decided to call their 1968 results “preliminary” when they found no statistically significant genetic influences at what should have been the study’s data-collection stop-point, and then kept looking for additional diagnosed parents after 1968 in the hope of finding enough schizophrenia spectrum diagnoses among their adopted-away biological offspring to achieve statistically significant results. Rosenthal and colleagues then decided to stop counting diagnoses at the point at which the comparison dipped below the .05 level of statistical significance, and published these results in 1971. Although such manoeuvring to arrive at desired conclusions may have been acceptable in 1960s-1970s psychiatry, it is now properly recognized as a classic p-hacking strategy (John and colleagues’ QRPs #2 and #4). In a different context, genetically oriented psychologist Stuart Ritchie criticized a p-hacking strategy of “not setting the sample size beforehand,” which allows “researchers to continue collecting data and testing it, collecting data and testing it, again and again in an open-ended way until they get their desired p < 0.05.” This describes to a T the p-hacking strategy Rosenthal, Kety, Wender and colleagues employed in this still-cited “landmark” schizophrenia adoption study.

Lynn DeLisi echoed mainstream psychiatric “turning point” descriptions of these adoption studies in her 2017 popular work 100 Questions & Answers About Schizophrenia: Painful Minds:

“The turning point in schizophrenia research, and probably the most important data collection and results of the 20th century in this field, came from the carefully planned and executed adoption studies of Seymour Kety and David Rosenthal….These data…turned the corner for support for ‘nurture’ to support for primarily ‘nature.’”

The reality is that the Danish-American schizophrenia adoption studies are perhaps the longest-running example of p-hacked research ever seen in the behavioural sciences, and as such should not survive the replication crisis. The schizophrenia genetics story unravels even more.

Final thoughts

Contrary to the consensus opinion in psychiatry and other behavioural sciences, there exists no scientifically acceptable evidence that schizophrenia or psychosis have an underlying genetic component, which has implications for other areas of human behaviour. When it is finally acknowledged that “schizophrenia” is not “genetic” or a “disease,” society will part ways with genetic diversions and inappropriate medical approaches, and will instead focus on environmental causes, non-medical interventions, and prevention with a necessary political-change component. People might still wish to take prescribed drugs, but the process should be completely transparent, including informing people of all potential side effects and withdrawal problems. No one should be told they are being prescribed drugs to treat a genetically caused schizophrenia brain disease—not the “patient,” not the family, not anyone.

Because family, social, cultural, religious, educational, geographical, and political environments together play a powerful role in shaping human behaviour, attention should be focused away from people’s brains and genes, and toward aspects of the environment that on the one hand help protect, nurture, and empower people, and on the other hand can psychologically harm people.

In his 1980 book The IQ Game, sociologist Howard Taylor described IQ-genetic researchers’ “use of assumptions that are implausible as well as arbitrary to arrive at some numerical value for the genetic heritability of human IQ scores on the grounds that no heritability calculations could be made without benefit of such assumptions”. Modifying Taylor’s description to fit schizophrenia genetic research and claims, in Schizophrenia and Genetics I describe the use of false or questionable assumptions, questionable research practices, and the promotion of spurious or non-causative gene association claims to produce schizophrenia heritability estimates and gene discovery claims, because no such estimates or claims could be made without engaging in such practices.

This is the schizophrenia game. It has been played for over a century, and it’s time to stop. A world without schizophrenia genes will do just fine.

****

The author thanks Mike Jones for helpful feedback on an earlier draft of this article.